0710.1634/ms.tex
1: 
2: \documentclass[8pt,preprint2]{aastex}
3: 
4: \shorttitle{Time Delays in SDSS~J1004+4112}
5: \shortauthors{Fohlmeister et al.}
6: 
7: \begin{document}
8: 
9: \title{The Rewards of Patience: An 822 Day Time Delay in the Gravitational Lens SDSS~J1004+4112}
10:  \author{J. Fohlmeister\altaffilmark{1},
11:    C. S. Kochanek\altaffilmark{2}, E. E. Falco\altaffilmark{3},
12:    C. W.  Morgan\altaffilmark{2,4},
13:    \and J. Wambsganss\altaffilmark{1}
14: }
15: 
16:    \altaffiltext{1}{Astronomisches Rechen-Institut, Zentrum f\"ur
17:               Astronomie der Universit\"at Heidelberg, M\"onchhofstr. 12-14,
18:               69120 Heidelberg, Germany} 
19:    \altaffiltext{2}{Department of Astronomy, Ohio State University, 140 West
20:               18th Avenue, Columbus, OH 43210}
21:    \altaffiltext{3}{Smithsonian Astrophysical Observatory, FLWO, P.O. Box
22:               97, Amado, AZ 85645} 
23:    \altaffiltext{4}{Department of Physics, United States Naval Academy, 572C Holloway Road,
24:               Annapolis, MD 21402}
25: 
26: \begin{abstract}
27: We present 107 new epochs of optical monitoring data for the four brightest
28: images of the gravitational lens SDSS J1004+4112 observed between October 2006
29: and June 2007.  Combining this data with the previously obtained light curves,
30: we determine the time delays between images A, B and C. We confirm our
31: previous measurement finding that A leads B by $\Delta t_{BA}=40.6\pm1.8$~days,
32: and find that image C leads image A by $\Delta\tau_{CA}=821.6\pm2.1$ days. The
33: lower limit on the remaining delay is that image D lags image A by 
34: $\Delta\tau_{AD}>1250$ days.  Based on the microlensing of images A and B
35: we estimate that the accretion disk size at a rest wavelength of 2300\AA\ is 
36: $10^{14.8\pm0.3}$~cm for a disk inclination of $\cos{i}=1/2$, which is consistent with
37: the microlensing disk size-black hole mass correlation function given our
38: estimate of the black hole mass from the MgII line width of $\log M_{BH}/M_\odot=8.44\pm0.14$. 
39: The long delays allow us to fill in the seasonal gaps and assemble a continuous,
40: densely sampled light curve spanning 5.7 years whose variability implies a
41: structure function with a logarithmic slope of $\gamma = 0.35\pm0.02$. 
42: As C is the leading image, sharp features in the C light curve can be 
43: intensively studied 2.3 years later in the A/B pair, potentially allowing
44: detailed reverberation mapping studies of a quasar at minimal cost. 
45: \end{abstract}
46: 
47: 
48:    \keywords{cosmology: observations -- 
49:              gravitational lensing --
50:                 quasars: individual: (SDSS J1004+4112)
51:                }
52: 
53: \section{Introduction}
54: 
55: The quasar SDSS~J1004+4112 at $z_s=1.734$ is split into five images by an intervening
56: galaxy cluster at $z_l=0.68$ \cite{inada,inada2,oguri}. With a maximum image separation of $14\farcs62$, it is
57: a rare example of a quasar gravitationally lensed by a cluster \cite{wambsganss,inada3}. 
58: One of the most interesting applications of this system is to use the time delays between
59: the lensed images to study the structure of the cluster.  If we assume the Hubble 
60: constant is known, then the delays break the primary model degeneracy of lensing studies
61: (the ``mass sheet degeneracy''), and the delay ratios constrain the structure even if
62: the Hubble constant is unknown.  After its discovery, several groups modeled the 
63: expected time delays in SDSS J1004+4112 and their dependence on the mean mass profile of 
64: the cluster \cite{kawano,oguri,williams}.   When we measured the shortest delay in the
65: system, between images A and B, we found a longer delay than predicted by the models
66: (Fohlmeister et al. 2007, hereafter Paper I) where the discrepancy probably arose 
67: because the models included the cD galaxy and the cluster halo but neglected 
68: the significant perturbations from the member galaxies.   As we measure the longer
69: delays, where the cluster potential should be relatively more important than for
70: the merging A/B image pair, we would not expect cluster substructures to play
71: as important a role.
72: 
73: We also expect this lens to have a fairly short time scale for microlensing variability
74: created by stars either in the intracluster medium or in galaxies near the images.
75: The internal velocities of a cluster are much higher than in a galaxy (700~km/s
76: versus 200~km/s), and  SDSS~J1004+4112's position on the sky is almost orthogonal 
77: to the CMB dipole (Kogut et al. 1993), giving the observer a projected motion on the lens
78: plane of almost 300~km/s.  In Paper I, we detected microlensing of the continuum emission of the 
79: A/B images in Paper I and there is also evidence for microlensing of the CIV broad line
80: \cite{richards,lamer,gomez}.  Once we have measured the time delays we can remove
81: the intrinsic quasar variability and use the microlensing variability to estimate the mean 
82: stellar mass and stellar surface density, the transverse velocities, and the 
83: structure of the quasar source \cite{gilmerino,mortonson,poin,morgan}. 
84: 
85: Finally, we note that SDSS~J1004+4112 could be an ideal laboratory for studying correlations
86: in the intrinsic variability of quasars.  With, image C leading images
87: A and B by 2.3 years, sharp variations in image C can be used to plan intensive monitoring
88: of images A and B to measure the response times as a function of wavelength
89: (e.g. Kaspi et al. 2007), with the additional advantage that the delay between A and B provides 
90: redundancies that
91: protect against weather, the Moon and the Sun.  The long delays between the images also
92: mean that seasonal gaps are completely filled, and we can examine the structure function
93: of the variability with a densely-sampled, gap-free light curve (modulo corrections for microlensing).
94: Such data generally do not exist, since most time variability data for quasars (other
95: than nearby reverberation mapping targets, e.g. Peterson et al. 2004) have very sparse sampling (e.g. Hawkins 2007
96: on long time scales for a small number of objects or Vanden Berk et al. 2004 on shorter time scales
97: for many objects).  
98: 
99: In Paper I \cite{fohli} we presented three years of optical monitoring data for the four 
100: brightest images of SDSS J1004+4112 spanning 1000 days from December 2003 to June 2006. 
101: The fifth quasar image, E, is too faint to be detected in our observations. We measured
102: the time delay between the A and B image pair to be $\Delta\tau_{BA}=38.4\pm2.0$ days. 
103: While larger separation lenses tend to have longer time delays, for these two images the 
104: propagation time difference is small, because they form a close image pair ($3\farcs8$) 
105: from the source lying close to a fold caustic.  For the more widely separated
106: C and D images we could only estimate lower limits on the delays of 560 and 800
107: days relative to image B and A.  In this paper we present the 107 new optical monitoring epochs 
108: for the 2006/2007 season in \S2.  When combined with our previous data we have light
109: curves spanning 1250 days that allow us to measure the AC delay in \S3.  In \S4
110: we use the microlensing variability of the A/B images to measure the size of the
111: quasar accretion disk, and in \S5 we measure the structure function of the intrinsic
112: variability.  We discuss the future prospects for exploiting this system in \S6.    
113: 
114: \begin{figure*}[t]
115:    \centering
116:    \includegraphics[bb= 30 100 520 700, width=10cm,angle=0,clip]{f1.ps}
117:    \caption{
118:    Light curves of the A, B, C and D images of the quasar SDSS J1004+4112 from 
119:    December 2003 to June 2007. Images C and D have been offset by 0.3
120:    and 1.0 mag, respectively, in order to avoid overlap. We present a running 
121:    average of one point every 5 days averaged over $\pm7$ days to emphasize
122:    trends and to avoid confusion by noise.   
123:    \label{lcurve}} 
124: \end{figure*}
125: 
126: 
127: 
128: \section{Data}
129: 
130: We monitored SDSS~J1004+4112 in the r-band during the 2006-2007 season using the Fred 
131: Lawrence Whipple Observatory (FLWO) 1.2m telescope on Mount Hopkins and the 
132: MDM 2.4m Hiltner Telescope on Kitt Peak.  The FLWO observations were obtained with 
133: Keplercam (0\farcs672 pixels) and the MDM observations with  RETROCAM \cite{morgan1}
134: (0\farcs259 pixels).  The data reduction was carried out as described in Paper I.
135: We continued to use the same five stars to set the PSF model and the flux 
136: scale of each epoch and verified that these flux standards continue to show
137: no variability. Table~1 presents the photometry for the four images in the
138: 2006-2007 season. 
139: 
140: In Figure \ref{lcurve} we present the resulting light curves for images A to D for 
141: the period from December 2003 to June 2007.  The average sampling rate during the 
142: 2006/2007 season is once every third day.  
143: The FLWO data are noisy, so for Figure \ref{lcurve} we show a running average 
144: of the data  (one point every five days averaged over $\pm$ 7 days) to emphasize
145: the long term trends.  Images C and D are offset by 0.3 and 1.0 mag, respectively, 
146: so that they do not overlap with image B in the third and fourth season.  During
147: this season, images A and B faded by approximately 0.4 mag with a prominent feature
148: near the middle of the season, image C was relatively constant and image D 
149: brightened by about 0.4 mag.  For the full four seasons, A and B have faded by
150: approximately 1~mag, C has remained relatively constant and D has brightened
151: by about 1.5~mag. 
152: 
153: \section{The Time Delay}
154: 
155: For the determination of the time delay, we use the methods described
156: in Paper I. Our first step with the new data was to remeasure the A/B
157: delay. The fourth season shows a nice feature with maxima in images
158: A and B near days 4120 and 4080 respectively, followed by a roughly 100 day 
159: decline to minima at 4220 (A) and 4180 (B) days. 
160: With the dispersion method \cite{pelt1,pelt2}
161: we measure the delay between A and B to be $\Delta t_{BA}=40.1\pm3.5$~days.
162: For the Kochanek et al. (2006) polynomial method we used 
163: polynomial orders of $N_{src}=20$, $40$, $60$ and $80$ for the
164: source and $N_\mu=1$, $2$, $3$ and $4$ for the microlensing 
165: variability and derived the final estimate using the Bayesian
166: weighting of these cases described in Poindexter et al. (2007).
167: We found delays of $40.6\pm1.8$, $40.1\pm1.8$ and $39.8\pm1.8$
168: (68\% confidence regions) depending on whether we weighted the changes 
169: in the number of parameters using the Bayesian information criterion (which strongly
170: penalizes extra parameters), the Akaike information criterion (which
171: weakly penalizes extra parameters) or no penalty for extra parameters.
172: These are consistent with our result from Paper I of $\Delta\tau_{BA}=38.4\pm2.0$ 
173: days, but are somewhat more conservative in their treatment of the
174: parameterization and the role of microlensing. 
175: 
176: In Paper I we derived a lower limit on the BC delay of $\Delta\tau_{CB}>560$ days
177: and suggested, based on some similarities between the third season for A/B with
178: the first season for C, that a delay of order 700 days was plausible but 
179: statistically too weak to claim as a measurement.  We now see that the
180: feature in the second season for image C strongly matches the feature
181: we observe in the new season for A and B.  Using the dispersion 
182: spectra method (Pelt et al. 1994, 1996), we find $\Delta\tau_{CA}=822\pm7$ days
183: and  $\Delta\tau_{CB}=780\pm6$ days where the CA delay is slightly less
184: accurate because the CA overlap is slightly less than the CB overlap due
185: to the alignment of the light curves relative to the seasonal gaps.  The
186: three delays are mutually consistent since 
187: $\Delta\tau_{CB}=\Delta\tau_{CA}-\Delta\tau_{BA}=782\pm7$ days.  For
188: the polynomial method analysis we simultaneously fit A, B and C
189: holding the A/B delay fixed to 40.6~days to find CA delays of 
190: $821.6\pm2.1$, $823.0\pm2.1$ and $820.2\pm2.1$~days for the
191: three weighting methods, respectively.  
192: Image D should lag the other three images, and we see no feature in the
193: light curve of image D that can be matched to the first season of images
194: A/B.   The lower limit on the time delay between images A and D is
195: now $\Delta\tau_{DA}>1250$ days (3.4 years).
196: 
197: We modeled the lens using the same approach as in Paper I, where we include
198: the central cD galaxy, and NFW halo for the cluster dark matter and 12
199: pseudo-Jaffe models corresponding to cluster
200: galaxies (we added an extra component at $(x,y)=(31\farcs0,4\farcs0)$ relative
201: to quasar image A in an effort to reduce the overall shear).  The 
202: fits were carried out using {\it lensmodel} (Keeton 2001) and while
203: adequate they are not satisfactory -- it is very difficult to find
204: solutions with no additional quasar images created by
205: the galaxies, and checking for the extra images makes the procedure
206: extraordinarily slow.  At present we lack the ability to model this
207: system in detail (including uncertainties) at the precision of the
208: constraints, while simplified models that ignore the galaxies are
209: incapable of fitting the data at all.  The model predicts an AD delay 
210: of order 2000~days (5.5 years), which is consistent with our current 
211: lower bound.
212: 
213: \section{Microlensing and the Size of the Quasar Accretion Disk}
214: 
215: The residuals of the A and B light curves (see Fig. \ref{delay}) clearly 
216: indicate that microlensing is present.   After correcting for the time delay,
217: the mean magnitude differences 
218: between A and B for the four seasons are $0.460 \pm 0.005$, $0.283\pm0.007$,
219: $0.339\pm0.005$ and $0.381\pm0.007$~mag.  For the two seasons overlapping
220: with C we find mean magnitude differences, seasonal gradients and second derivatives of 
221: $0.590\pm0.010$~mag, $-0.04\pm0.02$~mag/year and $0.29\pm0.09$~mag/year$^2$
222: for C relative to A and $0.368\pm0.005$~mag, $0.05\pm0.01$~mag/year and
223: $0.18\pm0.04$~mag/year$^2$ for B relative to A.  Fig.~\ref{delay} shows
224: the superposition of the phased A, B and C light curves and the differences
225: between them that are the signature of microlensing. 
226: 
227: We modeled the microlensing for images A/B using the Bayesian Monte Carlo
228: method of Kochanek (2004).  We used the microlensing parameters of our
229: (adequate) lens model, with convergence $\kappa$ and shear $\gamma$ 
230: values of $\kappa=0.48$ and $\gamma=0.57$ for A and 
231: $\kappa=0.47$ and $\gamma=0.39$ for B.  We allowed the surface
232: density in stars $\kappa_*$ to vary from 10\% to 100\% of $\kappa$
233: increments of 10\%.  We used a microlens mass function with 
234: $dn/dM \propto M^{-1.3}$ with a dynamic range in mass of a factor
235: of 50 that approximates the Galactic disk mass function of Gould (2000).  
236: We generated $4096 \times 4096$ pixel magnification patterns with 
237: an outer scale of $20\langle R_E\rangle$ where $\langle R_E\rangle$
238: is the Einstein radius at the mean stellar mass $\langle M\rangle$.
239: We modeled the disk as a face-on, thin disk \cite{shakura}
240:  neglecting the central temperature depression and
241: relativistic effects.  We measure the disk size $R_\lambda$ as the point
242: where the disk temperature matches the rest-frame energy of
243: our monitoring band, $k T_\lambda = hc/\lambda$, where 
244: $\lambda\simeq 2300$\AA\ for the r-band at the source redshift
245: (see Morgan et al. 2007).
246: The half-light radius $R_{1/2}=2.44R_\lambda$ should be used to
247: compare to any other disk model, since Mortonson et al. (2005)
248: have shown that the half-light radius depends little on the
249: surface brightness profile of the model.  We made four realizations
250: of each of the 10 microlensing models and drew $2\times 10^5$ trial
251: light curves for each of the 40 cases so that we would have a
252: reasonable statistical sampling of light curves that fit the 
253: data well.  We found that  
254: \begin{equation}
255:     R_{2300\AA} = 10^{14.8\pm 0.3} {\hbox{cm} \over h_{70} \sqrt{cos i}}
256: \end{equation}
257: for a disk inclination angle $i$,
258: whether or not we use a prior on the mean microlens mass of
259:  $0.1M_\odot < \langle M\rangle < M_\odot$.  
260: 
261: From the MgII emission line width/black hole mass calibration of 
262: Kollmeier et al.  (2006), the spectrum of image C from Richards 
263: et al. (2004), and a magnification-corrected HST $I$-band magnitude 
264: of $20.9\pm0.4$, we estimate a black hole mass of 
265: $\log M_{BH}/M_\odot =8.4\pm0.2$.  Fig.~\ref{disk} compares the 
266: disk size estimate to the characteristic scales of such a black hole.
267: 
268: \begin{figure}[t]
269:    \centering
270:    \includegraphics[width=7.5cm,angle=0,clip]{f2.ps}
271: 
272:    \caption{
273:     The image A, B and C light curves in their overlap region
274:     after shifting by the time delays. The data are binned in one week
275:     intervals. The lower box shows the residual magnitudes
276:     shifted by the offset between the images, revealing microlensing
277:     variability of order 0.15 mag.
278:     The light curve of image B was chosen to have constant flux because it
279:     has the most overlap with the over two.  
280:    \label{delay}} 
281: \end{figure}
282: 
283: \begin{figure}[t]
284:    \centering
285:    \includegraphics[width=7.5cm,angle=0,clip]{f3.ps}
286: 
287:    \caption{ Probability distribution for the accretion disk size at 2300\AA\
288:       assuming the mean disk inclination ($\cos(i)=1/2$).  The
289:       solid (dashed) curves show the distribution without (with) the prior on the
290:       average microlens mass.  The vertical lines show the Schwarzschild radius
291:       and the expected size for thin disk theory assuming the black hole mass
292:       estimated from the MgII emission line width.  The expectation from thin
293:       disk theory assumes the disk is radiating at Eddington ($L/L_E=1$) with
294:       efficiency $\eta=0.1$ where $L=\eta \dot{M} c^2$.
295:    \label{disk}} 
296: \end{figure}
297: 
298: \section{The Structure Function}
299: 
300: The quasar structure function can be used as a tool to characterize  quasar
301: variability independent of short-timescale monitoring gaps and to compare
302: with theoretical models of quasar variability (e.g. Kawaguchi et al. 1998).  
303: The structure function 
304:    \begin{equation}
305:      S(\tau) =\sqrt{{1\over N({\tau})} \sum\limits_{i<j} [m(t_{j})-m(t_{i})]^{2}}
306:    \end{equation}
307: is the variance in the magnitude as a function of the time $\tau=t_j-t_i$ between 
308: measurements where $m(t_{j})$ and $N({\tau})$ is the number of epochs at that 
309: time lag.  For SDSS J1004+4112 we can determine the structure function over a 
310: moderate time range and with a dense sampling rate and no seasonal gaps if we
311: use the time-delay corrected quasar light curves for images B and C.  These 
312: cover a time-baseline of 2065 days (5.7 years) in the observers frame, corresponding to a maximum
313: rest-frame time lag at $z_s=1.734$ of 755 days. For the very different behavior of the image
314: D light curve, which could not yet be time-delay connected to the other images, we
315: compute the structure function independently for rest frame time lags up to
316: 470 days. As in Vanden Berk et al. 2004 we fit the form of the structure function with a
317: power law. The value for the power law index $\gamma = 0.35\pm0.02$
318: for the combined image B and C light curves is consistent with that derived
319: for the SDSS quasar sample. For image D we find a similar slope of
320: $\gamma = 0.39\pm0.03$, as expected from the light curve. 
321: Time-delay
322: connecting the image A, B and C lightcurves by subtracting the estimated
323: microlensing variability in the overlap region of the lightcurves gives a
324: restframe record of the intrinisic quasar variability over 500 days. The slope
325: of the structure function for the source light curve $\gamma_{s} = 0.45\pm0.03$ is steeper than for the
326: observed non-microlensing corrected curves.  
327: 
328: \section{Summary and Conclusions}
329: 
330: We present a fourth season of monitoring data for the four bright images of the
331: five image gravitational lens system SDSS J1004+4112.  We confirm our previous
332: estimate for the time delay between the merging A/B pair, finding that B leads
333: A by $40.6\pm1.8$ days.  We measure the delay for image C for the first time,
334: finding that it leads image A by $821.6\pm2.1$~days.  We note that this is
335: nearly twice the longest previously measured delay (the 417 day delay in 
336: Q0957+561 \cite{schild, kundic}).  We find a lower
337: bound that D lags A by more than approximately 1250~days.  Our current mass model predicts 
338: that D lags A by approximately 2000 days, which is consistent with the present limit.
339: The fractional uncertainties
340: in the AB delay are still dominated by sampling and microlensing, while the
341: fractional uncertainties in the AC delay are dominated by cosmic variance due
342: to density fluctuations along the line of sight rather than our measurement
343: uncertainties of 0.3\% (e.g. Barkana 1996).  
344: 
345: A detailed model of this system, including the constraints from the multiply
346: imaged, higher redshift arcs (Sharon et al. 2005), the X-ray measurements
347: \cite{ota,lamer}
348: and a detailed understanding of the uncertainties will
349: be a challenge.   We lack a completely satisfactory model for the system
350: at present, in the sense that the modeling process is extraordinarily slow 
351: due to the ability of the gravitational potentials associated with the 
352: cluster member galaxies to generate additional but undetected images of the quasar,
353: making it impossible to carry out a reliable model survey.  The record of models
354: for this system is discouraging.  As we noted in Paper I, all three model
355: studies (Oguri et al. 2004; Williams \& Saha 2004;  
356: Kawano \& Oguri 2006) generically predicted shorter AB delays than the observed
357: 40 days, and that this could be plausibly explained by the absence of 
358: substructure (i.e. galaxies) in the potential models.  The longer AB-C
359: and AB-D delays should be less sensitive to substructure.  Oguri et al.
360: (2004) do not include an estimate of the AB-C delays and have A-D 
361: delays consistent with our present limits.  The range of B-C delays
362: in Williams \& Saha (2004) is consistent with our measurement of 820
363: days, but they predict AD delays shorter than our current lower bound
364: of 1250~days.  Kawano \& Oguri (2006) predict a range for the longer 
365: delays over a broad range of mass distributions, none of which match
366: our delays in detail.  However, models with sufficiently long C-B delays
367: generally have C-D delays long enough to agree with our present limits.
368: 
369: Based on our present mass model we used the microlensing between the A
370: and B images to make an estimate of the size of the quasar accretion disk
371: at 2300\AA\ in the quasar rest frame.  If we convert this to the expected
372: size at 2500\AA\ assuming the $R_\lambda \propto \lambda^{4/3}$ scaling
373: for a thin disk and assume the mean disk inclination $\cos (i)=1/2$ the scale 
374: on which the disk temperature matches the photon energy is $R_{2500\AA}=10^{15.0\pm0.3}$~cm.
375: Comparisons to other disk models should use the half-light radius which is $2.44$ times larger.
376: Based on the quasar MgII emission line width we estimate that the black hole 
377: mass is 
378: $10^{8.4\pm0.2} M_\odot$. For this mass, the microlensing accretion disk size-black
379: hole mass correlation found by Morgan et al. (2007) predicts that $R_{2500\AA}=10^{15.3}$~cm,
380: which is in broad agreement with the measurement.  Further observations, the
381: inclusion of additional images, and monitoring in multiple bands should
382: improve these measurements and potentially allow us to determine the mean
383: surface density in stars near the images $\kappa_*$ and their average mass
384: $\langle M\rangle$.  Similarly, the ability to construct continuous light
385: curves of the intrinsic variability and to use image C to provide early
386: warning of sharp flux changes that can then be intensively monitored in
387: images A and B may make this system a good candidate for applying 
388: reverberation mapping techniques to a massive, luminous quasar.  At
389: present, we already see that the system has a structure function 
390: typical of quasars.  
391: 
392: 
393: 
394: \begin{acknowledgements}
395:  We thank all the participating observers at both the Harvard-Smithsonian
396:  Center for Astrophysics and the MDM Observatory for their support of these
397:  observations.  This work is also based on observations obtained with the MDM
398:  2.4m Hiltner and 1.3m McGraw-Hill telescopes, which are owned and
399:  operated by a consortium consisting of Columbia University, Dartmouth
400:  College, the University of Michigan, the Ohio State University and
401:  Ohio University. We thank N.F. Bate for valuable comments and encouragements.
402:  We also acknowledge support by the European Community's Sixth
403:  Framework Marie Curie Research Training Network Programme, Contract
404:  No. MRTN-CT-2004-505183 ``ANGLES".
405: \end{acknowledgements}
406: 
407: %----------------------------------------------------------------
408: 
409: 
410: 
411: \begin{thebibliography}{}
412: 
413:  \bibitem[Barkana 1996]{barkana} Barkana, R., 1996, ApJ, 468, 17
414: 
415:  \bibitem[Fohlmeister et al. 2007]{fohli} Fohlmeister, J. et al. 2007, ApJ, 662, 62
416: 
417:  \bibitem[Gil-Merino et al. 2005]{gilmerino} Gil-Merino, R., Wambsganss,  J.,
418:  Goicoechea, L.J., Lewis, G.F. 2005, A\&A, 432, 83
419: 
420:   \bibitem[G\'omez-\'Alvarez et al. 2006]{gomez} G\'omez-\'Alvarez, P., Mediavilla, E., Mu\~noz, J.A.,
421:   Arribas, S., S\'anchez, S.F., Oscoz, A., Prada, F., \& Serra-Ricart, M.,
422:   2006, ApJL, 645, 5
423: 
424:  \bibitem[Gould 2000]{gould} Gould, A. 2000, ApJ, 535, 928
425: 
426:  \bibitem[Hawkins 2007]{hawkins} Hawkins, M. R. S. 2007, A\&A, 462, 581
427: 
428:   \bibitem[Inada et al. 2003]{inada} Inada, N. et al. 2003, Nature, 426, 810
429:   \bibitem[Inada et al. 2005]{inada2} Inada, N. et al. 2005, PASJ, 57, L7
430:   \bibitem[Inada et al. 2006]{inada3} Inada, N. et al. 2006, ApJ, 653, 97
431: 
432:   \bibitem[Kaspi et al. 2007]{kaspi} Kaspi, S., Brandt, W.N., Maoz, D.,
433:   Netzer, H., Schneider, DP.,  \& Shemmer, O. 2007, ApJ, 659, 997
434: 
435:   \bibitem[Kawaguchi et al. 1998]{kawaguchi} Kawaguchi, T., Mineshige, S.,
436:   Umemura, M., \& Turner, E.L. 1998, ApJ, 504, 671 
437: 
438:   \bibitem[Kawano \& Oguri 2006]{kawano} Kawano, Y., \& Oguri, M. 2006, PASJ, 58, 271 
439: 
440:   \bibitem[Keeton 2001]{keeton} Keeton, C. R. 2001, astro-ph/0102340
441: 
442:   \bibitem[Kochanek et al. 2004]{kochanek0} Kochanek, C. S. 2004, ApJ, 605, 58 
443: 
444:   \bibitem[Kochanek et al. 2006]{kochanek} Kochanek, C. S., Schneider, P., Wambsganss, J. 2006, Gravitational
445:   Lensing: Strong, Weak \& Micro, Proceedings of the 33rd Saas-Fe Advanced Course, G. Meylan,
446:   P. Jetzer \& P. North, eds. (Springer Verlag: Berlin)
447: 
448:   \bibitem[Kochanek et al. 2006b]{kochanek2} Kochanek, C. S., Morgan, N. D., Falco,
449:   E. E., McLeod, B. A., Winn, J. N., Dembicky, J., \& Ketzeback,
450:   B. 2006, ApJ, 640, 47
451: 
452:  \bibitem[Kogut et al. 2003]{kogut} Kogut, A., et al., 1993, ApJ, 419, 1
453: 
454:  \bibitem[Kollmeier et al. 2006]{kollmeier} Kollmeier, J. A. et al. 2006, ApJ,
455:  648, 128
456: 
457:  \bibitem[Kundic et al. 1997]{kundic} Kundic, T., Turner, E.L., Colley, W.N., Gott, J.R.,
458:    Rhoads, J.E., Wang, Y., Bergeron, L.E., Gloria, K.A., Long, D.C., Malhotra, S., \&
459:    Wambsganss, J., 1997, ApJ, 482, 75
460: 
461:  \bibitem[Lamer et al. 2006]{lamer} Lamer, G., Schwope, A., Wisotzki, L.,  Christensen, L., 2006, A\&A, 454, 493
462: 
463:  \bibitem[Morgan et al. 2005]{morgan1} Morgan, C.W., Byard, P.L., Depoy, D.L., Derwent, M.,
464:      Kochanek, C.S., Marshal, J.L., O'Brien, T.P., \& Pogge, R.W., 2005, AJ, 129, 2504
465:  \bibitem[Morgan et al. 2007]{morgan} Morgan, C.W., Kochanek, C.S., Morgan,
466:  N.D., Falco, E.E., 2007, astro-ph/0707.0305
467: 
468:  \bibitem[Mortonson et al. 2005]{mortonson} Mortonson, M.J., Schechter, P.L. \& Wambsganss, J. 2005, ApJ, 628, 594
469: 
470:  \bibitem[Oguri et al. 2004]{oguri} Oguri, M., et al. 2004, ApJ, 605, 78
471: 
472:   \bibitem[Ota et al. 2006]{ota} Ota, N., et al. 2006, ApJ, 647, 215
473: 
474:   \bibitem[Pelt et al. 1994]{pelt1} Pelt, J., Hoff, W., Kayser, R., Refsdal, S.,
475:        \& Schramm, T. 1994, A\&A, 286,775
476:   \bibitem[Pelt et al. 1996]{pelt2} Pelt, J., Kayser, R., Refsdal, S.,
477:        \& Schramm, T. 1996, A\&A, 305, 97
478:  
479:   \bibitem[Peterson et al. 2004]{peterson} Peterson, B.M., et al. 2004, ApJ,
480:   613, 682
481: 
482:  \bibitem[Poindexter et al. 2007]{poin} Poindexter, S., Morgan, N.D., Kochanek, C.S., 2007, astro-ph/0707.0003
483: 
484:   \bibitem[Richards et al. 2004]{richards} Richards, G. T., et al. 2004, ApJ, 610, 679
485: 
486:   \bibitem[Schild \& Thomson 1995]{schild} Schild, R.E., \& Thomson, D.J., 1995, Aj, 109, 1070
487: 
488:   \bibitem[Shakura \& Syunyaev 1973]{shakura} Shakura, N. I., Syunyaev, R. A. 1973,
489:   A\&A, 24, 337
490: 
491:   \bibitem[Sharon et al. 2005]{sharon} Sharon, K. et al. 2005, ApJ, 629, 73
492:  
493:   \bibitem[Vanden Berk et al. 2004]{vandenberk} Vanden Berk, D.E., et al., 2004, ApJ, 601, 692
494: 
495:   \bibitem[Wambsganss 2003]{wambsganss} Wambsganss, J., 2003, Nature, 426, 781
496: 
497:   \bibitem[Williams 2004]{williams} Williams, L. L. R. \& Saha, P. 2004 , AJ,
498:   128, 2631
499: 
500: \end{thebibliography}
501: 
502: \clearpage
503: \begin{deluxetable}{cccccccc}
504: \rotate
505: \tablewidth{0pt}
506: \tablecaption{Light Curves for SDSS J1004+4112 \tablenotemark{*}}
507: \tablehead{
508: \colhead{HJD}           & \colhead{$\chi^{2}/N_{dof}$}      &
509: \colhead{Image A}       & \colhead{Image B}  &
510: \colhead{Image C}       & \colhead{Image D}  &
511: \colhead{Observatory} & \colhead{Detector}}
512: \startdata
513: \input tab1.tex
514: %\input tab1_stub.tex
515: \tableline
516: \enddata
517: \tablecomments{The Heliocentric Julian Days (HJD) column gives the date of the observation
518:   relative to HJD$=2450000$.  
519:   The $\chi^2/N_{dof}$ column indicates how well our photometric model fit the imaging
520:   data.  When $\chi^2>N_{dof}$ we rescale the photometric errors presented in this
521:   Table by $(\chi^2/N_{dof})^{1/2}$ before carrying out the time delay analysis to
522:   reduce the weight of images that were fit poorly. 
523:    The image magnitudes are relative to the comparison stars 
524:   (see text). The magnitudes enclosed in parentheses are not used in the time 
525:   delay estimates. } 
526: \end{deluxetable}					
527: 
528: \clearpage
529: 
530: 
531: \end{document}
532: